Home Search Collections Journals About Contact us My IOPscience

The width of margins in radiotherapy treatment plans

This content has been downloaded from IOPscience. Please scroll down to see the full text.

2000 Phys. Med. Biol. 45 3331

(http://iopscience.iop.org/0031-9155/45/11/315)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 193.140.151.85
This content was downloaded on 07/11/2014 at 06:28

Please note that terms and conditions apply.

Phys. Med. Biol. 45 (2000) 3331–3342. Printed in the UK PII: S0031-9155(00)13236-1

The width of margins in radiotherapy treatment plans

Alan L McKenzie†, Marcel van Herk‡ and Ben Mijnheer‡

† Bristol Oncology Centre, Horfield Road, Bristol BS2 8ED, UK
‡ Radiotherapy Department, The Netherlands Cancer Institute, Plesmanlaan 121,
1066 CX Amsterdam, The Netherlands
E-mail: alan.mckenzie@ubht.swest.nhs.uk, portal@nki.nl and bmijnh@nki.nl

Received 12 April 2000, in final form 21 July 2000

Abstract. Publication of ICRU Reports 50 and 62 has highlighted the need to devise protocols for
the process of drawing the planning target volume (PTV) around the clinical target volume (CTV).
The margin surrounding the CTV should be wide enough to account for all geometric errors so
that no part of the CTV accumulates a dose less than, for instance, 95% of that prescribed. One
approach to the problem has been to draw a margin around the CTV delineated at the treatment
preparation stage which is sufficiently wide that the mean position of the CTV will be encompassed
in a specific percentage of cases, for example 90%. This accounts for the systematic errors. A
further margin is then drawn to account for random set-up and organ-motion uncertainties during
treatment.
The width of this second margin has previously been shown to be 1.64(σ − σp ). Here
σ , a vector quantity, is the standard deviation which results from convolving the penumbra spread
function of standard deviation σp with the Gaussian distributions of the daily positional uncertainties
of organ motion and set-up error. However, it is shown in this paper that the calculation should
take into account the beam configuration of the treatment plan. In a typical coplanar multibeam
plan, usually in the transverse plane, any given edge of the target volume is normally defined by
a single beam or two parallel and opposed beams. However, because of the presence of the other
beams, the effect of the blurring of the edge-defining beam(s) is reduced, which changes the value
of the required margin to β(σ − σp ) where, for example, β can be as low as 1.04 in the transverse
plane of a three-beam plan.
The width of the required margins is calculated for up to six beams and presented in a table. It
is shown that, while the table was derived using an idealized plan of equally weighted plane beams
irradiating a spherical target, it is also valid for non-uniform beam weightings, wedged-beam plans,
target volumes of general shape and intensity-modulated radiotherapy (IMRT).

1. Introduction

People who produce computer treatment plans have for many years been aware of the effect of
beam configuration on the penumbral width of the isodoses surrounding the target volume. For
example, in conformal therapy with coplanar beams in the transverse plane, a greater margin
between the beam edge and the target volume is needed in the superior–inferior direction
than in the transverse plane. However, these margins are determined only by the measured
penumbra widths of the treatment beams, and take no account of the uncertainty introduced by
organ motion or set-up errors in daily treatment. This paper combines the above experience
of treatment planners with the model proposed by van Herk et al (2000) which accounts for
systematic and random errors, and shows how the beam configuration determines the margin
widths required to accommodate the daily treatment uncertainties of organ motion and set-up
errors.

0031-9155/00/113331+12$30.00 © 2000 IOP Publishing Ltd 3331

3332 A L McKenzie et al

The publication of ICRU Reports 50 and 62 (ICRU 1993, 1999) has highlighted the need
to devise protocols for the process of drawing the planning target volume (PTV) around the
clinical target volume (CTV). In particular, the problem of accounting for geometric errors has
attracted significant attention and generated several different approaches (see, for example,
Leong 1987, Austin-Seymour et al 1995, Jones et al 1995, Stroom et al 1999, Ekberg et al
1998, Roach et al 1994, Mageras et al 1999).
It is convenient to account separately for the effects of systematic and random errors on
the width of the CTV–PTV margins. Work by van Herk and colleagues (2000) suggests a
system for accounting for both systematic and random errors in the radiotherapy treatment
process, and they show how to determine the margin width required for a single beam or a
fully conforming dose distribution delivering dose within specified limits such as the minimum
of 95% of prescription dose within the CTV. This is consistent with the lower limit of 95%
of prescribed dose for the PTV recommended by ICRU 50 (ICRU 1993). A limitation of the
work by van Herk is that idealized conformation—where a single beam is ‘moulded’ around
the complete 3D surface of the target volume—is, of course, impossible in practice. This paper
builds upon the work of van Herk and colleagues and suggests how to decide a margin width
for the typical multibeam configurations found in treatment plans.

2. Theory

2.1. Systematic errors

Errors in delineating the CTV at the treatment preparation stage will be propagated throughout
the complete fractionation schedule and are therefore systematic errors which should be
accommodated by appropriately wide margins. (A consistent error in setting up the patient
arising, for example, from a laser which is misaligned throughout the treatment, is also a
systematic error.)
A margin drawn around the CTV to account for systematic errors produces an extended
CTV. The width of this margin determines the probability that the extended CTV encompasses
the ‘real’, average position of the CTV—that is, the position of the CTV as it would appear
if there were no delineating or imaging errors, organ motion or uncertainties in the doctor’s
identification of the CTV. Changes in the shape and size of the CTV also introduce positional
uncertainty in the boundary of the CTV which contribute to the required margin width.
Although errors such as organ motion occur randomly, the position of the CTV when it is
imaged on the CT scanner will be ‘frozen’ for the whole of the treatment because the treatment
plan will be based on this position. That is why such an error occurring during treatment prepa-
ration is called a systematic error. However, subsequent organ motion during the course of treat-
ment also needs to be accounted for, and this is discussed in the next section on random errors.
If the standard deviations of the individual systematic uncertainties can be estimated, they
can be combined into a standard deviation which we shall call . This standard deviation
is a measure of how far the image of the CTV produced at the treatment preparation stage is
likely to be displaced from the mean, ‘real’ position. Note that the standard deviations of both
systematic and random errors are vectors with components, for example, in the three cardinal
directions (superior–inferior, lateral and antero-posterior).
Assuming a three-dimensional Gaussian distribution of displacement, van Herk and
colleagues (2000) show that a margin of 2.5 drawn around the CTV produced at the treatment
preparation stage will encompass the mean, ‘real’ position of the CTV in 90% of patients. This
is illustrated in figure 1 where the use of ellipsoidal surfaces accommodates different values
of the components of  in different directions.
 Width of margins in radiotherapy treatment plans 3333

Figure 1. The doctor’s delineation of the CTV results in a systematic error . In 90% of cases the
mean position of the CTV will be entirely encompassed by drawing a margin 2.5 wide around
the delineated CTV.

It should be noted that while Gaussian distributions are usually acceptable approximations
for distributions of displacement, organ motion resulting from breathing is an exception and so
margins for breathing error should be added linearly and not combined as a Gaussian function
(McKenzie 2000).

2.2. Random errors

Throughout treatment, random errors are introduced in the daily set-up process in combination
with variations in the target position caused by organ motion. These random errors are
accounted for by adding a further margin to the extended CTV. The objective is to have
such a large margin between the beam edge and the extended CTV that, over the course of the
fractionation schedule, the accumulated dose is for instance no less than 95% of the prescribed
dose, anywhere within the extended CTV.
This objective is achieved by supposing the treatment beam to be ‘blurred’ to an extent
which depends upon the size of the daily random errors. These errors include the motion of the
target, but it is easier to think of the target as stationary and, instead, to consider the blurring
of the beam as being caused by motion of the beam only (Leong 1987).

2.3. Margin width for a single beam

The unblurred beam profile may itself be thought of as a convolution of a top-hat profile with a
pencil beam of Gaussian width (standard deviation) σp . The blurring operation then becomes
a convolution of the result with a Gaussian function of width σrandom which represents the daily
random uncertainties of set-up and target motion. That result can also be achieved directly by
convolution of the top-hat profile with a Gaussian of standard deviation σ which includes the
random uncertainties of set-up and target motion as well as the pencil beam, that is

 
σ = σp2 + σrandom
2
.
The blurred penumbra which results from this convolution is an error function. The resulting
profile of ‘blurred dose’ can also be understood as the relative proportion of the total duration
3334 A L McKenzie et al

Figure 2. The blurred beam may be represented by a convolution of the random error σ and a step
function. The error σ includes the width of the unblurred penumbra.

for which a given point is exposed to the non-zero region of the top-hat profile. This, in turn,
depends upon the relative probability of the beam edge being at the given point, which is
represented by the Gaussian distribution. It is convenient later to refer to the probability of
accumulated dose represented by the error function as a percentage ‘level of blurred dose’.
As the top-hat function is more or less flat, the 50% level of blurred dose clearly coincides
with the beam edge, while the 95% level of blurred dose, under the Gaussian approximation,
lies inside the beam at a distance of 1.64σ from the beam edge. This follows from tables of
the error function, and is illustrated in figure 2. The constant is of course an approximation
depending upon how well the contributing uncertainties are represented by Gaussian functions.
As shown in figure 3, following van Herk et al , we enlarge the extended CTV with a margin
of size 1.64(σ − σp ) to achieve the PTV and then perform standard treatment planning. In that
standard treatment planning process the beam width will be increased by 1.64σp beyond the
PTV, so that the 95% isodose as displayed on the treatment planning system will encompass the
PTV. (Notice that the 95% isodose displayed on the treatment planning system corresponds
to the 95% dose level for the unblurred beam, because the treatment planning system, of
course, does not contain information about blurring caused by random errors.) As a result, the
extended CTV lies at a distance of 1.64σ from the beam edge and so receives at least 95% of
the prescribed dose as intended.
In passing, it is of interest to note that, in the limit of random errors being zero, σ becomes
σp , and so the PTV becomes identical with the extended CTV. In this case, in the standard
treatment planning process, the beam width is again increased by 1.64σp beyond the PTV
(which is now also the extended CTV) and so the 95% isodose as displayed on the treatment
planning system will in this case encompass the extended CTV. Since for this example we
have surmised no random errors, the 95% isodose displayed on the treatment planning system
is a genuine measure of the dose delivered in this instance, so that the extended CTV again
receives at least 95% of the prescribed dose as intended.
It is important to distinguish the margin derived to accommodate daily random errors
from that needed to account for systematic errors. The former is derived using essentially a
one-dimensional approach because, for any beam edge, the blurring is in one direction (normal
to the beam edge). However, drawing a volume which encompasses the CTV and accounts for
systematic errors in the position of the CTV is a three-dimensional problem. From the above
discussion and that in section 2.1, it will be seen that the overall margin which should be drawn
between the CTV and the PTV is 2.5 + 1.64(σ − σp ) (see figure 3).
 Width of margins in radiotherapy treatment plans 3335

Figure 3. This figure illustrates the derivation of the CTV–PTV margin width for single-beam
irradiation. The appropriate margin is derived by considering that in the treatment planning process
the 95% isodose level due to unblurred beams is aligned with the PTV, while in reality the cumulative
dose distribution is blurred due to random variations, and the 95% isodose level of the cumulative
dose distribution should be aligned with the extended CTV.

It is the purpose of this paper to suggest how to determine margin widths for the more
realistic case of multiple-beam treatment plans, where the combination of beams modifies the
above expression for margin width.

2.4. Combination of beams

The solution of this problem is complex in the general case, and the problem could be
approached more easily by performing the treatment plan using blurred beams (i.e. beams
that are convolved with appropriate Gaussian functions describing the random variations).
However, commercial systems do not offer this facility, and so it is necessary to use a
pragmatic, geometric approach. The simplifying assumptions will be highlighted in the
following derivation. We consider only coplanar combinations of beams, and, for ease of
illustration, we assume that the gantry axis is aligned with the long axis of the patient, i.e. in the
cranio-caudal (superior–inferior) direction. We also assume initially that all of the irradiating
beams are plain rather than wedged, and consider later the size of the errors introduced with
this assumption.
The gantry axis is perpendicular to the plane of figure 4 which is drawn through the centre
of a spherical target volume (the extended CTV). Beam #1 is one of three blurred beams which
are assumed to irradiate the target with equal dose weight at the isocentre. Clearly, at the
point where the beam edge approaches the target volume most closely, the surface of the target
volume must be parallel to the beam edge, and a line can be drawn perpendicularly to both
surfaces at the point of closest approach. Since, at this point, beam #1 approaches the target
volume more closely than the other two beams, it is the edge of beam #1 which defines the
3336 A L McKenzie et al

Figure 4. The margin in a multibeam plan is defined by the beam edge which lies closest to the
target volume, because the gradient of the other beams at this point is much smaller.

Figure 5. This figure illustrates that, in a three-beam plan, it is the 85% dose level of a single
blurred beam defining the margin which corresponds to the 95% level of the total blurred dose.

margin at this point. For illustration, the profile of dose along the perpendicular is indicated
to show how the edge of beam #1 defines the margin.
Figure 5 shows this dose profile in more detail. In this figure the contribution of each
of the three beams is shown. In this example, we assume that none of the beams are parallel
and opposed to any other. Since the treatment plan will have been designed to produce a
homogeneous dose distribution across the target, the sum of beams #2 and #3 at the edge of
beam #1 must also be comparatively flat. That is, at the edge of beam #1, the gradient due to
beam #1 is much larger than the gradient due to the other two beams and so the 95% level of
blurred dose is fully determined by beam #1. Consequently, the effect of adding more beams
is reducing the distance of the 95% level of the blurred dose to the beam edge.
In the scale for the combined plots at the left side of the plot in figure 5, the sum of the
contributions of beams #2 and #3 may be seen to be 67%. On the right of the plot a scale
is drawn for beam #1 alone. Clearly, the 95% level of blurred dose for the total of the three
beams corresponds in this case to the 85% level of blurred dose for beam #1. Tables of the
error function show that the 85% level of blurred dose is situated 1.04σ from the beam edge.
 Width of margins in radiotherapy treatment plans 3337

Where none of the beams are parallel and opposed to another, and where a total of n beams
contribute with equal weights to the target irradiation, the 95% level of blurred dose from all n
beams corresponds to the following level of blurred dose for the beam that delimits the target
(beam #1 in the example):
level of blurred dose (no parallel opposed beams) = (100 − 5n)%. (1)
If the target volume in this example is not spherical, then the axis along which the dose profile
is drawn may not pass through the prescription point. This would mean that the 100% level of
the dose profile of the delimiting beam (beam #1) would not correspond to a beam weighting
of 1/n, which is the assumption upon which equation (1) is founded. However, as we shall see,
in practice, for most clinical target volumes, the error introduced by non-spherical volumes is
acceptably small.
With parallel and opposed beams, the combination of the two opposed beams effectively
doubles the gradient. Furthermore, in the case of coplanar beams (i.e. in the transverse plane
as usual), the edges of all of the beams together define the margin in the direction normal
to the plane (usually in the superior–inferior direction). Equation (1) may be modified to
accommodate these three situations:
level of blurred dose = (100 − 5n/p)% (2)
where n is the total number of beams irradiating the target, p = 1 (delineating beam is not
parallel and opposed), p = 2 (delineating beam is parallel and opposed) or p = n (when
considering dose profile in the superior–inferior direction).
Figure 6 shows the profile of dose which would be calculated for unblurred beams in
a treatment plan of the three-beam example above. The 95% isodose in the treatment plan
corresponds to the 85% dose level of the unblurred beam #1, which lies 1.04σp distant from
the beam edge. Hence, the 95% isodose in the treatment plan lies a distance 1.04(σ − σp )
from the position of the 95% level of blurred dose for all three beams (figure 6). Since the
95% level of blurred dose is to be matched with the boundary of the extended CTV, the overall
width of the margin required is 2.5 + 1.04(σ − σp ).
As a numerical example, consider the case of σp = 3.2 mm and σrandom = 3.3 mm.
Then σ = (3.22 + 3.32 )1/2 = 4.6 mm. In the case of a single-beam irradiation, the
extended CTV–PTV margin is 1.64(4.6 − 3.2) = 2.3 mm, and for a three-beam plan it
is 1.04(4.6 − 3.2) = 1.5 mm. Where organ motion is more apparent, for example where
σrandom = 10 mm, the extended CTV–PTV margin for a single-beam plan and a three-beam
plan would be 1.64(10.5 − 3.2) and 1.04(10.5 − 3.2) = 12.0 mm and 7.6 mm respectively.

2.5. General formula for margin widths

In general, for the coefficient of the term (σ − σp ) we use the term β. Values of β are listed in
table 1 for treatment plans containing up to six beams. This table has been constructed using
equation (2) and tables of the error function in the same way as for the three-field example
above. The general formula to use for the overall width of the margin between the CTV and
the PTV is then
CTV → PTVmargin width = 2.5 + β(σ − σp ). (3)
This equation is identical in form to that of van Herk and colleagues (2000). However, in
this case, the function of β is subtly different. In van Herk’s paper, the formula applies to a
single beam or a perfectly conformal dose distribution, and the choice of β depends upon the
level of blurred dose with which it is desired to treat the CTV. In equation (3), it is assumed
3338 A L McKenzie et al

Figure 6. This figure illustrates the derivation of the CTV–PTV margin width in the case of a
three-beam plan.
Table 1. In practice, the approximations made in constructing this table warrant only two digits
instead of three. However, three have been quoted because it is easier to see the derivation using
error function tables.
Value of β for penumbra of beam being considered
n (total number
of beams Transverse plane Transverse plane Cranio-caudal
irradiating (beam not parallel (beam parallel and direction (superior–
target) and opposed) opposed) inferior)
1 1.64 1.64
2 1.28 1.64 1.64
3 1.04 1.44 1.64
4 0.84 1.28 1.64
5 0.67 1.15 1.64
6 0.52 1.04 1.64

that we wish to treat the CTV to at least 95% of the prescription dose and the choice of β is
determined in table 1 by the beam configuration on the treatment plan.
Table 1 was calculated assuming that only one beam (or two parallel and opposed beams)
in the transverse plane would be tangential at any point. However, as the number of treatment
beams irradiating the target increases, so an increasing number of beams will be virtually
tangential to the surface of the target at any point on its circumference. Hence, these beams no
longer contribute a flat component to the ‘background’ profile of dose along a perpendicular to
the target volume. As a result, table 1 will underestimate β at high beam numbers. Instead, β
will, in practice, approach a lower limiting value. However, computer simulation of this effect
shows that, for up to six beams, table 1 is satisfactory.
 Width of margins in radiotherapy treatment plans 3339

Figure 7. This figure illustrates the application of table 1 to wedged beams. Effectively, the wedge
changes the beam weights at the anterior and posterior side. However, the effect of a typical wedge
on the required value of β is very small.

3. Discussion

We have calculated table 1 with the aim of delivering no less than 95% of the prescribed
dose anywhere within the extended CTV. The question is, is this a reasonable level of dose
to aim for? ICRU Report 50 (ICRU 1993) recommends that the dose throughout the PTV
should not drop below 95%, but we have shown in figure 3 that it is only the 95% isodose of
the unblurred beam which matches the PTV, and that the blurred dose will be less than 95%
outside of the extended CTV. The dose to the CTV will generally be higher than the 95% level
on the boundary of the extended CTV. ICRU Report 50 does not discuss how high the dose
to the CTV needs to be, but it can be inferred that a dose of close to 95% will occur where
the PTV–CTV margin is small. It is within this spirit that the level of 95% was chosen for
calculating table 1.
A numerical example will serve to put the use of equation (3) and table 1 in perspective.
Using the values of 3.2 mm and 3.3 mm assumed earlier for σp and σrandom respectively, as
well as a value of 3.3 mm for , the difference in changing β from 1.64 to 0.52 amounts
to a change in margin from 10.6 mm to 9.0 mm. Where organ motion is more apparent, for
example σrandom = 10 mm, the corresponding change in margin width is from 20.3 mm to
12.1 mm.
It is important to remember the assumptions used in constructing table 1, and it is
appropriate to ask for an estimate of the error inherent in these assumptions. A specific
example will illustrate the approximate size of the error incurred in assuming that the beams
are unwedged, while in reality beams are often wedged.
Consider the example in figure 7 which shows a three-field treatment plan containing
a single anterior beam and two lateral, parallel and opposed wedged beams. Each beam is
assumed to have the same weighting so that each contributes 33% of the dose to the prescription
point. From table 1, the value of β for the penumbra formed by the two lateral beams at the
anterior and posterior limit of the target volume is 1.44. In determining this value of β, the
assumption was that the contribution to the dose of the anterior beam at both the anterior and
posterior edges of the target volume is 33%.
However, for a 6 MV beam and a target volume extending 10 cm in the antero-posterior
direction, a more realistic estimate is that the contribution to the dose of the anterior beam
3340 A L McKenzie et al

might be 43% at the anterior edge and 25% at the posterior edge. Hence, for a uniform
distribution of dose throughout the target volume, the total contribution to the dose of the
lateral beams must be 57% and 75% respectively, which is achieved by a suitable choice of
wedge angles. At the anterior penumbra, for each of the two lateral beams, the level of blurred
dose corresponding to the 95% level of prescription dose will be 100 − 5/0.57 = 91.2% and,
at the posterior penumbra, the level of blurred dose is 100 − 5/0.75 = 93.3%. Tables of the
error function give values for β of 1.36 and 1.50 respectively for these percentages. This means
that the anterior margin could be slightly smaller than the tabulated one, whereas the posterior
margin should be larger. (For a representative value of 5 mm for (σ − σp ), the tabulated value
underestimates the required posterior margin by 0.3 mm.)
In order to determine the effect of this underestimate for the posterior margin, notice that
using the tabulated value of 1.44 for β would correspond to making the edge of the target
volume coincident with the 92.5% level of blurred dose for the individual lateral beams. (This
is because a value of 1.44 for β is derived assuming a nominal weighting of 67% for the sum of
the two lateral beams. This corresponds to 100 − 5/0.67 = 92.5% for the individual beams.)
However, rather than the nominal 67%, the lateral beams contribute 75% of the prescription
dose at the posterior edge of the beam, and so the 92.5% level of blurred dose contributes only
75% × 92.5% to the total dose which, together with the 25% contribution from the anterior
beam, amounts to a total of 94.4% of the prescription dose. This underdosing by 0.6% is likely
to be clinically acceptable.
A similar analysis may be made for the error in assuming equal beam weightings instead
of the actual weightings used in a given plan.
Although a spherical target volume was used to illustrate the derivation of table 1, the
values of β in the table are not particularly sensitive to the shape of the target volume, as
shown by the following argument. If the outline of the target volume in figure 4 were not
circular, the line perpendicular to both the beam edge and the surface of the non-spherical
target volume would in general no longer go through the isocentre. Therefore the level of the
‘background’ contribution to the dose profile from the other two beams would be different
from the 67% which was assumed in deriving table 1 for this beam configuration.
However, for a non-spherical target volume, in the worst case the perpendicular line will
be shifted either to the most anterior or the most posterior border of the target volume. In that
case, the values which were used for the background contribution in the wedged example can
be used again, and similar reasoning will once more demonstrate that the values in table 1 are
acceptable in practice.
The above discussion prompts the question of whether table 1 is still valid in the case of
intensity modulated radiotherapy (IMRT). The answer is that table 1 may still be applied for
the range of IMRT plans most commonly encountered in clinical practice. To illustrate this,
consider a five-beam IMRT treatment of the prostate with dose escalation to the inner part of
the target volume provided by a higher energy fluence in the central parts of the beams. As
before, the margin width is determined at the point of closest approach of any given beam
(beam #1) to the outer edge of the target volume. In this example, the point will be caught
in the ‘crossfire’ of the other four beams and, if the beam profiles were uniform instead of
modulated, table 1 would apply directly as before. In the case of beam modulation, the point
may be irradiated by one or more of the central portions of the crossfire beams. This would
reduce the relative contribution of beam #1 to the dose at the margin of the target volume. For
example, if the escalated dose in the centre of the target is 80 Gy and the outer dose is 60 Gy,
the relative contribution of beam #1 will fall from 20% in the case of homogeneous beams to
no lower than 15.8% in the case where the point is caught in the crossfire of the high-fluence
part of all four remaining beams. However, we have shown in the discussion above that larger
 Width of margins in radiotherapy treatment plans 3341

changes in the contribution from beam #1 (ranging from 25% to 43%) do not significantly
alter the value of β. The same argument can be made for a wide range of modulated beams,
and so table 1 may be used for IMRT.
More sophisticated IMRT planning techniques have been suggested which are capable of
accommodating geometric (and other) uncertainties in an approach which obviates the simple
method of drawing margins. For example, Löf et al (1998) have proposed an elegant stochastic
optimization technique which accounts for internal organ motion. This process generates a
beam profile which not only extends beyond the target volume but is also modulated in a way
which overcompensates the fluence near the beam edges. The same authors present an adaptive
feedback algorithm using interfractional measured set-up errors to modify the beam profile to
correct for such set-up errors. Clearly, when such techniques are used, the approach of simply
drawing margins to account for set-up errors and random organ motion is redundant.
van Herk and colleagues suggested using the approximation γ σ  for the term β(σ − σp )
in equation (3) where σ  represents the combined standard deviation of the organ motion and
set-up errors without including the penumbra width. For a single beam where β = 1.64
they give γ = 0.7 for values of σ and σp likely to be encountered in clinical practice. Bel
et al (1996) convolved realistic treatment-plan dose distributions with Gaussian distributions
of translational and rotational error to determine the change of position of the 95% isodose.
This enabled them to predict the necessary margin to account for such errors. For multibeam
treatment plans they found that translational errors in the clinical range (0–5 mm) generally
required smaller margins than predicted by the γ = 0.7 rule which applies to the single-beam
case. That observation is in agreement with this work.
Other approaches such as that of coverage probability used by Stroom et al (1999) have
also produced formulae similar in form to equation (3), although the concept of coverage
probability is not an exact parallel to that used in this work.

4. Conclusion

In conclusion, while it is well known that margin widths generally need to be different
in different directions to accommodate isotropies such as organ motion, it should also be
appreciated that the number and configuration of the treatment beams will introduce an
additional directional dependence for the margin width required. This paper shows that, once
the standard deviations in the different directions have been determined, the required widths
of these margins in different directions can be calculated using equation (3) with the values of
β given in table 1.

References

Austin-Semour M, Kalet I, McDonald J, Kromhout-Schiro S, Jacky J, Hummel S and Unger J 1995 Three dimensional
planning target volumes: a model and a software tool Int. J. Radiat. Oncol. Biol. Phys. 33 1073–80
Bel A, van Herk M and Lebesque J V 1996 Target margins for random geometrical uncertainties in conformal
radiotherapy Med. Phys. 23 1537–45
Ekberg L, Holmberg O, Wittgern L, Bjelkengren G and Landberg T 1998 What margins should be added to the clinical
target volume in radiotherapy treatment planning for lung cancer Radiother. Oncol. 48 71–7
ICRU (International Commission on Radiation Units and Measurements) 1993 Prescribing, recording and reporting
photon beam therapy ICRU Report 50 (Bethesda, MD: ICRU)
——1999 Prescribing, recording and reporting photon beam therapy (supplement to ICRU Report 50) ICRU Report
62 (Bethesda, MD: ICRU)
Jones D, Hafermann M D, Rieke J W and Vermeulen S S 1995 An estimate of the margin required when defining
blocks around the prostate Br. J. Radiol. 68 740–6
3342 A L McKenzie et al

Leong J 1987 Implementation of random positioning error in computerized radiation treatment planning systems as
a result of fractionation Phys. Med. Biol. 32 327–34
Löf J, Lind B K and Brahme A 1998 An adaptive control algorithm for optimization of intensity modulated radiotherapy
considering uncertainties in beam profiles, patient set-up and internal organ motion Phys. Med. Biol. 43 1605–28
Mageras G S, Fuks Z, Leibel S A, Ling C C, Zelefsky M J, Kooy H M, van Herk M and Kutcher G J 1999 Computerized
design of target margins for treatment uncertainties in conformal radiotherapy Int. J. Radiat. Oncol. Biol. Phys.
43 437–45
McKenzie A L 2000 How should breathing motion be combined with other errors when drawing margins around
clinical target volumes? Br. J. Radiol. 73 973–7
Roach M, Pickett B, Rosenthal S A, Verhey L and Phillips T L 1994 Defining treatment margins for six field conformal
irradiation of localized prostate cancer Int. J. Radiat. Oncol. Biol. Phys. 28 267–75
Stroom J C, de Boer H C, Huizenga H and Visser A G 1999 Inclusion of geometrical uncertainties in radiotherapy
treatment planning by means of coverage probability Int. J. Radiat. Oncol. Biol. Phys. 43 905–19
van Herk M, Remeijer P, Rasch C and Lebesque J V 2000 The probability of correct target dosage: dose-population
histograms for deriving treatment margins in radiotherapy Int. J. Radiat. Oncol. Biol. Phys. 47 1121–35